Hi all,

at the moment I work on a flow simulation of a water pump. The impeller rotates with 2000rpm. Could someone tell me which turbulence model would be the appropriate one for such an application.

Thanks in advance

Marcio

Marcio,

Any of the standard k-e turbulence models (when used with an appropriately refined mesh) should get you answers for the bulk flow features that are within 10% of reality. It's been a while since I did a pump inlet, but I'd be tempted to refine the grid enough to get y+ values near 1 and either use a k-omega model or an SST model. Either would most likely improve accuracy a little.

Alton

Thanks for your reply,

but I 've still a question concerning the turbulence model. If I understand you right then a turbulence model without wall functions is more precise than the high reynolds models. Is this the case???

You need to understand how wall functions are derived to answer this question. A review of a viscous flow book (good one is by F. White) will help. Wall functions provide an estimate of boundary layer shape below the mesh resolution you have used. If you look at the background, you will quickly realize that the theory was developed for flat plate flows, i.e. equilibrium, zero pressure gradient. For such flows, by plotting u+ versus y+, the turbulent boundary layers collapse into a single relation. However, there is no such behavior for adverse pressure gradient flows, especially the strongly adverse pressure gradient flows in pumps and compressors.

So the answer to your question - a low reynolds number turbulence model is certainly more precise for your adverse pressure gradient case than a high reynolds numbers that incorporates wall functions that do not match the physics of your case.

I still think Marcio needs an answer on what to do to get reasonable results (without the need for a mesh fit for a supercomputer) within a reasonable time. I would go for the Alton Reich answer and go for basic things like using wall functions, get 10+ nodes in the boundary layer and check the difference between a standard k-eps, SST, and maybe an RSM model to see what the range in your answer really is.